Classification of microarrays; synergistic effects between normalization, gene selection and machine learning

Abstract

BackgroundMachine learning is a powerful approach for describing and predicting classes in microarray data. Although several comparative studies have investigated the relative performance of various machine learning methods, these often do not account for the fact that performance (e.g. error rate) is a result of a series of analysis steps of which the most important are data normalization, gene selection and machine learning.ResultsIn this study, we used seven previously published cancer-related microarray data sets to compare the effects on classification performance of five normalization methods, three gene selection methods with 21 different numbers of selected genes and eight machine learning methods. Performance in term of error rate was rigorously estimated by repeatedly employing a double cross validation approach. Since performance varies greatly between data sets, we devised an analysis method that first compares methods within individual data sets and then visualizes the comparisons across data sets. We discovered both well performing individual methods and synergies between different methods.ConclusionSupport Vector Machines with a radial basis kernel, linear kernel or polynomial kernel of degree 2 all performed consistently well across data sets. We show that there is a synergistic relationship between these methods and gene selection based on the T-test and the selection of a relatively high number of genes. Also, we find that these methods benefit significantly from using normalized data, although it is hard to draw general conclusions about the relative performance of different normalization procedures.

Gene Expression Profiles for Predicting Metastasis in Breast Cancer: A Cross-Study Comparison of Classification Methods

“…For example, one study comparing the performance of eight different classification methods showed that NNET and SVMs in general perform better than the other six methods for predicting outcome in eight different cancer microarray datasets [15]. Several studies confirm our finding of RF inferiority when using cross-validation [17, 18, 40]. Interestingly, a study conducting algorithm comparison on microarray gene expression based drug signatures showed that NNET and R-SVM had the best performance when tested in the most heterogeneous datasets [41].…”

“…However, few studies have systematically compared the predictive performance of such methods using microarray gene expression datasets on breast cancer. In their studies, method comparisons have been done within the same datasets by, for example, 10-fold cross-validation, leave-one-out cross-validation, or hold-out procedures [14–18], addressing prediction of relapse within a 5-year period [14, 16, 19], or molecular subtype classification [15]. Furthermore, even fewer studies have compared cross-study validation between classification methods within the field of breast cancer research.…”

Gene Expression Profiling of Bronchoalveolar Lavage Cells Preceding a Clinical Diagnosis of Chronic Lung Allograft Dysfunction

“…The points on the boundaries are called support vectors, while our optimal separating hyperplane is located in the middle of the margin. Literature suggests SVM classifiers have superior and robust performance in identifying predictive biomarkers in the setting of high-dimensional microarray gene expression data [14–18]. To overcome overfitting due to small number of arrays and large number of features, we also performed recursive feature elimination (SVM-RFE) algorithm [19] on the SVM to remove features with smallest ranking criterion, which corresponds to components of the SVM weight vector that are smallest in absolute value.…”

An Institutional Approach to Solar Fuels Research

“…[24,35,[46][47][48][49][50][51][52][53][54][55] Bio4Energy is led by Stellan Marklund and is funded with 20 million euros by the Swedish government for the initial 5-year period 2010-14. Further constellations at Umeå University that deal with forest feedstock optimization include Formas-financed Strong Research Environments BioImprove (2.5 million euros) and FuncFiber (2.5 million euros), [56][57][58][59][60][61][62][63] the Wallenberg-financed Conifer Genome Consortium (7.5 million euros), and the UPSC Berzelii Centre for Forest Biotechnology (10 million euros) [47,64,65] financed by the Swedish Research councils VR and Vinnova.…”